安徽省土壤湿度时空变化规律分析及遥感反演

为获取安徽省的土壤湿度时空信息,采用克里金法将站网实测多层土壤湿度数据插值为网格数据,分析其时空变化特征;进而建立遗传算法优化的BP(back propagation)神经网络模型进行土壤湿度反演。该模型以风云3B卫星的亮温数据为主要输入,训练后对该模型验证并进行预测。结果表明:安徽省土壤湿度月均值波动较频繁,淮北平原和大别山区较其他区域干燥;随着深度的增加,土壤湿度增大且季节和空间差异变小;所有分区平均模拟值与实测值的日序列相关性达到0. 605,均方根误差为0. 056 m~3/m~3,说明该模型能够较好地反演安徽省土壤湿度。

Spatiotemporal analysis and remote sensing retrieval of soil moisture across Anhui Province, China

To obtain the spatiotemporal characteristics of soil moisture in Anhui Province, the Kriging method was firstly used to interpolate the in-situ observed and multilayer soil moisture to gridded data. Then, the spatiotemporal variability of soil moisture across this region was analyzed. A Back Propagation(BP)neural network optimized by the genetic algorithm was established to retrieve the soil moisture using the brightness temperature measured by the Fengyun 3B satellite. The results show that soil moisture across Anhui Province shows high temporal fluctuations. And, the soil moisture in the Huaibei plain and the Dabie Mountains is lower than the other regions. As depth becomes deeper, soil moisture has a higher value with lower seasonal and horizontal variability. The correlation between retrieved and observed daily gridded values across the five sub-regions is 0. 605, while the corresponding root mean square error is 0. 056 m³/m³. Clearly, the proposed retrieval algorithm is able to capture the spatiotemporal variability of soil moisture in Anhui Province.